An intelligent watch data security transmission method and system based on artificial intelligence

Abstract

The application discloses an intelligent watch data security transmission method and system based on artificial intelligence, relates to the technical field of data transmission, and comprises the following steps: collecting multi-source type data through an intelligent watch sensor to form a type data set; determining a disclosure probability according to log access and public matching times of the type data, analyzing sensitivity based on the collection times, and summarizing and calculating a privacy score; analyzing the importance score of the type data according to a user request; determining a real-time score according to delay analysis of the data; introducing an entropy weight method to dynamically calculate a weight; and calculating a comprehensive sensitivity score.The method disclosed by the application dynamically adjusts the weight of each parameter through the entropy weight method and information entropy, so that the influence of each parameter on the comprehensive sensitivity score is enhanced, and the influence of parameters with weak distinguishing ability is reduced; and the introduction of the priority group distribution ratio ensures that the node weight reflects the distribution characteristics of the comprehensive sensitivity.

Description

Technical Field

[0001] This invention relates to the field of data transmission technology, and in particular to a method and system for secure data transmission from a smartwatch based on artificial intelligence. Background Technology

[0002] As an important device, smartwatches are not limited to providing basic functions such as time display, but have also become important data collection tools in fields such as personal health management, location tracking and behavior analysis. The multi-sensor architecture of smartwatches enables them to collect various types of data in real time, including but not limited to heart rate, steps, blood oxygen concentration, geographical location and environmental data. These large-scale, multi-source heterogeneous data provide data support for applications in medical care, sports analysis and behavioral data mining.

[0003] However, these data also face serious security and privacy threats during generation, transmission, and storage. If highly sensitive user data (such as heart rate and location information) is maliciously accessed or leaked, it may cause serious personal privacy violations. Although there have been some studies on optimizing data transmission in smartwatches in recent years, such as improvements in encryption for highly sensitive data and exploration of some path optimization algorithms, existing technologies still have the following shortcomings: First, existing solutions lack multi-dimensional dynamic analysis for data privacy protection and fail to comprehensively calculate the priority distribution of data from the perspectives of privacy, real-time performance, and importance synergy. This results in highly sensitive data not being adequately protected, while unnecessary encryption of low-sensitivity data consumes additional resources, leading to a decrease in transmission efficiency. Second, in terms of path selection, it is impossible to flexibly adjust based on the latency, load status, and distribution of sensitive data of dynamic links, making it difficult to balance network transmission efficiency and reliability. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method and system for secure data transmission in smartwatches based on artificial intelligence. Existing solutions lack multi-dimensional dynamic analysis for data privacy protection and fail to comprehensively calculate the priority distribution of data from the perspectives of privacy, real-time performance, and importance. This results in insufficient protection for highly sensitive data, while unnecessary encryption of low-sensitivity data consumes additional resources, leading to a decrease in transmission efficiency. Secondly, in terms of path selection, it is impossible to flexibly adjust based on the latency, load status, and distribution of sensitive data at nodes of dynamic links, making it difficult to balance network transmission efficiency and reliability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for secure data transmission from a smartwatch based on artificial intelligence, which includes collecting multi-source data of various types through the smartwatch's sensors to form a type dataset;

[0008] The probability of public access is determined based on the log access and public matching frequency of type data. Sensitivity is analyzed based on the number of collections, and a privacy score is calculated. The importance score of type data is analyzed based on user requests. The real-time score is determined based on the data latency analysis. The entropy weight method is introduced to dynamically calculate the weights and calculate the comprehensive sensitivity score.

[0009] Based on the comprehensive sensitivity score, the distribution of the data is analyzed to determine the skewness coefficient. The k-means clustering algorithm is used to initially group all the comprehensive sensitivity scores. The value of the initial cluster is determined by the elbow method. The dynamic weight is determined based on the distribution skewness coefficient and the comprehensive sensitivity score. The cluster values ​​are then weighted and optimized to obtain the optimized priority group. A graph model composed of data transmission nodes of smartwatch devices is used. The global weight is defined based on the data ratio of the priority group, and the edge weight is calculated. The Dijkstra algorithm is used to perform path selection decision to determine the optimal path.

[0010] The system is tiered based on priority groups, and candidate encryption algorithms are determined for each group. Data fragmentation is performed on the type data in the priority groups.

[0011] The encrypted data is segmented and stored for verification according to priority levels. An LSTM model is constructed to predict link attributes, and anomaly analysis is performed based on the prediction error.

[0012] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, the method of introducing entropy weighting to dynamically calculate weights and calculate a comprehensive sensitivity score includes:

[0013] The data access frequency data of the collected data types is statistically analyzed. The number of log accesses is determined based on the access logs of each data type. At the same time, the number of public matches of each data type is obtained from the user-published database, and the public probability is calculated.

[0014] Analyze the dynamic adjustment coefficients for potential sensitivity based on the total number of times each type of data is collected;

[0015] Privacy scores are calculated by combining potential sensitivity and public probability. In statistical data, the business function coverage of different sensor types is analyzed, and an audit index is defined by considering the number of user requests. Importance scores are then calculated.

[0016] For each type of data The real-time performance score is determined by analyzing the delay between the timestamp data and the current time.

[0017] The data is normalized based on its real-time score, importance score, and privacy score, and the entropy weight method is introduced to dynamically calculate the weights and calculate the comprehensive sensitivity score.

[0018] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, the following steps are included: determining dynamic weights based on the distribution skewness coefficient and comprehensive sensitivity score; weighted optimization of cluster values ​​to obtain optimized priority groups; a graph model composed of data transmission nodes of the smartwatch device; defining global weights based on the data proportions of the priority groups; calculating edge weights; and performing path selection decisions using the Dijkstra algorithm to determine the optimal path.

[0019] Distribution analysis is performed on the comprehensive sensitivity scores of different types of data to calculate the standard deviation of the comprehensive sensitivity distribution, and the skewness of the data is determined based on the skewness coefficient.

[0020] Based on the set of comprehensive sensitivity scores, the k-means clustering algorithm is used to initially group all comprehensive sensitivity scores, and the total within-group squared error is calculated. Based on the total within-group squared error and the initial group number, an elbow curve is plotted. The elbow point, i.e. the initial group number, is the position where the decrease in the total within-group squared error slows down significantly, and the value of the initial cluster is determined.

[0021] Based on the clustering results, the weighted silhouette coefficient is verified, the silhouette coefficient is calculated, and the weighted optimization is performed to obtain the optimized centroid number. The comprehensive sensitivity score data is then clustered into priority groups, and the center value of each group is attached.

[0022] Define smartwatch devices as device nodes, and set them as intermediate nodes and terminal nodes respectively according to the data transmission gateway and server. Calculate the distribution ratio based on the distribution of the type of data collected by the nodes in the priority group, and define the global weight of each node based on the clustered group data.

[0023] Calculate the edge weights based on the node's transmission links, the link transmission latency, and the global weight of the maximum load.

[0024] Set the optimal objective based on the edge weights, and select the path with the largest total weight as the optimal path;

[0025] The Dijkstra algorithm is used to perform path selection decisions. The highest priority node is selected from the adjacent nodes of the initial node. All edges connected to the node are traversed. For each edge, the path weight of the target node is updated according to the cumulative edge weight. If the cumulative weight is larger, the path source of the target node is updated. The target node is added to the queue. When the terminal node is dequeued, the iteration backtracks the path and generates the optimal path.

[0026] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, the step of classifying data based on priority groups, determining candidate encryption algorithms for each priority group, and performing data fragmentation processing on the data types in the priority groups includes:

[0027] Priority is ranked based on the average comprehensive sensitivity score of priority groups. The priority threshold is set by the sum of the historical average score and twice the standard deviation. Priority groups with average scores greater than the priority threshold are marked as high priority groups, and priority groups with average scores less than or equal to the priority threshold are marked as low priority groups.

[0028] The sets of candidate encryption algorithms are determined based on the high-priority group, the low-priority group, and the high-latency link, respectively.

[0029] The corresponding encryption algorithm is determined for each side link of the optimal path, and the data of the type in the priority group is processed by data fragmentation, where the fragment size u is dynamically adjusted based on the link load and bandwidth capacity.

[0030] Each fragment is encrypted according to the encryption algorithm of the path and its corresponding link.

[0031] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, the step of storing and verifying encrypted data fragments according to priority levels includes:

[0032] Differentiated chain storage rules are designed for encrypted data sharding, including full encrypted storage and private key management for high-priority data, and partial data transmission through cross-chain transfer protocol for low-priority data, with block-level verification of integrity.

[0033] For nodes with different storage priorities, the integrity of the uploaded data fragment chain is verified step by step.

[0034] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, the step of constructing an LSTM model to predict link attributes and performing anomaly analysis based on the prediction error includes:

[0035] A deep learning model based on LSTM neural network is constructed, including an input layer, an LSTM layer and an output layer. The input layer takes in a time series consisting of link transmission delay, bandwidth utilization, link packet loss rate and retransmission count. The LSTM layer extracts the temporal relationship of the time series and outputs the predicted value for the next time step through the output layer.

[0036] The model is trained using calibrated training set data. The cross-entropy loss function is selected to calculate the difference between the predicted class probability and the actual label. The Adam optimizer is used for gradient descent optimization to update the model parameters. If the model loss no longer decreases significantly during continuous iteration, the iteration stops and the model parameters are output.

[0037] Collect real link status data, calculate the predicted Euclidean distance based on the model output, and use it as the prediction error. Use the sum of the mean of historical prediction error values ​​and twice the standard deviation as the detection threshold. If the prediction error is greater than or equal to the detection threshold, it is judged as data anomaly, and the path selection is re-performed.

[0038] As a preferred embodiment of the AI-based smartwatch data security transmission method of the present invention, wherein: the step of collecting multi-source data through smartwatch sensors to form a type dataset includes,

[0039] Multi-source data is collected through smartwatch sensors, and each type of data is combined into a type dataset, where each type of data... This includes sensor type, data value, and timestamp.

[0040] Secondly, the present invention provides a smartwatch data security transmission system based on artificial intelligence, comprising,

[0041] The type data management module collects multi-source type data from smartwatch sensors and aggregates collection log information, including access logs and public match counts for each data type.

[0042] The comprehensive scoring module calculates a comprehensive sensitivity score for each type of data by combining privacy score, importance score, and real-time score.

[0043] The priority division module analyzes the distribution characteristics of the comprehensive sensitivity score, performs preliminary grouping based on the comprehensive sensitivity score, and optimizes the priority grouping and its group center value based on the distribution skewness coefficient.

[0044] The path planning module and the graph construction module apply Dijkstra's algorithm to select the path with the largest cumulative edge weight as the optimal data transmission path.

[0045] The hierarchical data storage module analyzes and encrypts data according to priority, performs data fragmentation, and stores data in a hierarchical manner.

[0046] The anomaly detection module constructs an LSTM neural network to extract temporal relationships and predict the link state attributes at the next time step, thereby performing anomaly detection.

[0047] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the data security transmission method for a smartwatch based on artificial intelligence as described in the first aspect of the present invention.

[0048] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the AI-based smartwatch data secure transmission method described in the first aspect of the present invention.

[0049] The beneficial effects of this invention are as follows: By using the entropy weight method, the weight of each parameter is dynamically adjusted through information entropy, thereby enhancing its influence on the overall sensitivity score and reducing the influence on parameters with weaker discriminative ability. The introduction of priority group distribution ratios ensures that the node weights reflect the distribution characteristics of overall sensitivity while avoiding the meaningless amplification of the influence of certain secondary sensor data due to the diversity of a single device node. By avoiding a one-size-fits-all equal allocation of resources during the transmission of data from different priority groups, the delay or risk exposure of highly sensitive data is avoided. At the same time, the path for transmitting low-sensitivity data at high frequencies is more inclined to performance optimization. The data fragmentation processing of type data in the priority group achieves lightweight and flexible data stream transmission, and the fragment size is dynamically adjusted according to the link load and bandwidth capacity. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart illustrating the AI-based smartwatch data security transmission method in Example 1.

[0052] Figure 2 This is a schematic diagram of the data security transmission system for a smartwatch based on artificial intelligence in Example 1. Detailed Implementation

[0053] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0054] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0056] Example 1, referring to Figures 1 to 2 This is the first embodiment of the present invention, which provides a method for secure data transmission in a smartwatch based on artificial intelligence, including the following steps:

[0057] S1 collects multi-source data through the smartwatch's sensors to form a type dataset;

[0058] Preferably, multi-source data of various types are collected through the sensors of a smartwatch to form a type dataset, including:

[0059] The smartwatch collects multi-source data (including heart rate (HR), steps (STEPS), geolocation (GPS), blood oxygen (SpO2), and user identity information) through its sensors. Each data type is then compiled into a categorized dataset. This includes sensor type, data value, and timestamp.

[0060] S2. Determine the public probability based on the log access and public matching times of type data, analyze the sensitivity based on the number of collections, and summarize and calculate the privacy score. Analyze the importance score of type data based on user requests, determine the real-time score based on data latency analysis, introduce the entropy weight method to dynamically calculate the weight, and calculate the comprehensive sensitivity score.

[0061] Preferably, the entropy weight method is introduced to dynamically calculate the weights and calculate the comprehensive sensitivity score, including:

[0062] The data access frequency data for each data type is collected statistically. This is determined by the number of log accesses for each data type, and the public probability is calculated based on the number of publicly available matches for each data type obtained from the user-published database. This probability is expressed as:

[0063] ;

[0064] in This indicates that the data collected by the i-th sensor The public probability, This indicates that the data collected by the i-th sensor The number of public matches, This indicates that the data collected by the i-th sensor Number of visits;

[0065] Based on the total number of data collections for each data type, the dynamic adjustment coefficient for potential sensitivity is analyzed and expressed as follows:

[0066] ;

[0067] in Indicates sensor type Sensitivity weights, where n represents the total number of types. Indicates sensor type Frequency of use;

[0068] The privacy score is calculated by combining potential sensitivity and probability of disclosure, and is expressed as:

[0069] ;

[0070] in Represents type data Privacy score;

[0071] In the statistical data, the business function coverage of different sensor types is considered, and an audit index is defined based on the number of user requests. An importance score is then calculated and expressed as follows:

[0072] ;

[0073] ;

[0074] in Indicates sensor type The business function coverage rate, where m represents the number of functions currently supported by the smartwatch. Represents a Boolean variable. Indicates sensor type If it is used by the j-th function, otherwise it is 0. Represents type data Importance score The audit index is indicated based on the sensor type. The ratio of the number of requests to the total number of requests used to request all data is used to determine the data.

[0075] For each type of data The real-time performance score is calculated based on the delay between the timestamp data and the current time, and is expressed as follows:

[0076] ;

[0077] in Represents type data Real-time score, Indicates the current time. Represents type data The collection timestamp, This indicates the maximum allowable delay, determined based on historical experience;

[0078] The data is normalized based on its real-time score, importance score, and privacy score. An entropy weighting method is then introduced to dynamically calculate the weights, resulting in a comprehensive sensitivity score, expressed as:

[0079] ;

[0080] ;

[0081] ;

[0082] in This represents the normalized weight of the q-th item in the real-time score, importance score, and privacy score. and Representing type data respectively The entropy value of the score of the qth item. This represents the value of the natural logarithm with respect to n, used in the calculation of standardized entropy. Represents type data The normalized value of the score for the q-th item, where n represents the total number of data types. Represents type data The overall sensitivity score.

[0083] Calculating the probability of public disclosure provides a clear basis for judging whether a certain type of data is over-disclosed, quantifies its risk, and, combined with the potential sensitivity adjustment coefficient, the synergistic effect of the two allows the privacy score to not only quantify the security risks caused by public access to data, but also dynamically analyze the additional threats that data intensity (total amount collected) may pose to privacy.

[0084] The introduction of importance scores not only plays a role in privacy and security decisions for sensor types, but also avoids one-sided analysis of sensor acquisition behavior based solely on sensitivity or privacy. In typical real-time monitoring scenarios with dynamic and high-frequency transmission, real-time scores can effectively balance data privacy and importance, and determine the strength of relevant restrictions on the time sensitivity dimension. This makes the calculation of comprehensive sensitivity scores more comprehensive and flexible, adaptable to dynamic multi-task data acquisition scenarios, while further preventing efficiency or security risks caused by high latency.

[0085] With the introduction of privacy, importance, and real-time scores, the entropy weighting method dynamically adjusts the weight of each parameter through information entropy, enhancing its influence on the overall sensitivity score while reducing the influence on parameters with weaker discriminative ability. This avoids decision-making errors that may occur when fixed weights are manually applied, potentially overlooking certain important scenarios. It optimizes the robustness of the dynamic data processing solution and, with the support of the overall score, achieves globally controllable, efficient transmission and encrypted closed-loop processing.

[0086] S3. Based on the comprehensive sensitivity score, perform distribution analysis on the type data to determine the distribution skewness coefficient. Use the k-means clustering algorithm to initially group all comprehensive sensitivity scores. Determine the value of the initial cluster using the elbow method. Determine the dynamic weight based on the distribution skewness coefficient and the comprehensive sensitivity score. Perform weighted optimization on the cluster values ​​to obtain the optimized priority group. Build a graph model composed of data transmission nodes from smartwatch devices. Define global weights based on the data proportions of the priority groups and calculate edge weights. Use the Dijkstra algorithm to perform path selection decisions and determine the optimal path.

[0087] Preferably, dynamic weights are determined based on the distribution skewness coefficient and comprehensive sensitivity score. The cluster values ​​are then weighted and optimized to obtain optimized priority groups. A graph model composed of data transmission nodes from smartwatch devices is used. Global weights are defined based on the data proportions of the priority groups, and edge weights are calculated. The Dijkstra algorithm is then used to perform path selection decisions to determine the optimal path, including...

[0088] Distribution analysis is performed on the comprehensive sensitivity scores of different types of data. The standard deviation of the comprehensive sensitivity distribution is calculated, and the skewness of the data is determined based on the skewness coefficient, as shown below:

[0089] ;

[0090] in The coefficient representing the skewness of the distribution of the comprehensive sensitivity score set. This represents the mean of the overall sensitivity score. This represents the standard deviation of the overall sensitivity score;

[0091] Based on the set of comprehensive sensitivity scores, the k-means clustering algorithm is used to initially group all comprehensive sensitivity scores, and the total within-group squared error is calculated. An elbow curve is plotted based on the total within-group squared error and the initial group number. The elbow point, i.e., the initial group number, is where the decrease in the total within-group squared error slows down significantly. The value of the initial cluster is then determined, expressed as:

[0092] ;

[0093] in This represents the total within-group squared error. Indicates the initial number of groups. This represents the i-th group. Denotes the centroid of the i-th group (i.e., all the centroids of that group). (mean)

[0094] Based on the clustering results, weighted silhouette coefficients are validated, the silhouette coefficients (measuring clustering quality) are calculated, and weighted optimization is performed to obtain the optimized centroid number. The comprehensive sensitivity score data are then clustered into priority groups, with the group center values ​​appended, as follows:

[0095] ;

[0096] ;

[0097] ;

[0098] in The silhouette coefficient represents the initial cluster value. express The average distance to other points within the same cluster. express The average distance to the nearest other clusters, The weighting factor represents the dynamic equilibrium. This represents the optimized number of centroids;

[0099] The smartwatch device is defined as a device node, and the data transmission gateway and server are respectively set as intermediate nodes and terminal nodes. Based on the distribution of the data collected by the nodes in priority groups, the distribution ratio is calculated. For each node, a global weight is defined based on the clustered data, expressed as:

[0100] ;

[0101] ;

[0102] in This represents the global weight of node v. and These represent the data proportions of the i-th and j-th priority groups corresponding to node v, respectively. The influence of multiple data types is controlled by these proportions to avoid weight inflation due to an excessive number of data types. This indicates that the data transmitted in node v belongs to a priority group. The number of data entries;

[0103] Based on the node's transmission link, and according to the global weight of the link transmission delay and maximum load, the edge weight is calculated and expressed as:

[0104] ;

[0105] ;

[0106] in This represents the link edge weight between the i-th node and the j-th node. This represents the link security weights of the i-th node and the j-th node. This represents the transmission delay of the link between the i-th node and the j-th node. This represents the maximum load of the link transmission between the i-th node and the j-th node. and Let represent the global weights of the i-th node and the j-th node, respectively;

[0107] The optimal objective is set based on the edge weights, and the path with the largest total weight is selected as the optimal path, expressed as:

[0108] ;

[0109] in Indicates priority group The optimal transmission path, Represents the weight of all edges in the transmission path. The summation is performed, and the path consists of multiple edges. Composition, each edge Each represents a link in the network. The link weight reflects the quality of its transmission conditions, and the total weight of the entire path is obtained by accumulating them.

[0110] The Dijkstra algorithm is used to perform path selection decisions. The highest priority node is selected from the adjacent nodes of the initial node. All edges connected to the node are traversed. For each edge, the path weight of the target node is updated according to the cumulative edge weight. If the cumulative weight is larger, the path source of the target node is updated. The target node is added to the queue. When the terminal node is dequeued, the iteration backtracks the path and generates the optimal path.

[0111] By measuring the standard deviation and skewness coefficient of the comprehensive sensitivity score set, the discreteness and distribution properties of the data are respectively measured. In subsequent cluster analysis based on the comprehensive sensitivity score, the skewness distribution can more accurately reflect whether the highly sensitive categories are prominent, which helps to guide the reasonable selection and initialization of the number of data cluster groups.

[0112] By combining clustering with elbow curves, the selection of the initial number of clusters is made more scientific and more adaptable. After the initial grouping, the silhouette coefficient measures the density of data points within each cluster and the separation between clusters to evaluate the clustering quality. The dynamic balancing weight factor optimizes the cluster centroid for the data distribution of each cluster. The segmentation of different priority groups can more reasonably reflect the differences between sensitivity scores. Finally, priority groups and their center values ​​of smartwatch data are generated.

[0113] By introducing the priority group distribution ratio, the node weights are ensured to reflect the distribution characteristics of comprehensive sensitivity, while avoiding the meaningless amplification of the impact of some minor sensor data due to the diversity of a single device node. The calculation of global weights lays the foundation for the allocation of link transmission edge weights, making the sensitivity of the data source more instructive for the selection of secure transmission paths in the entire network.

[0114] By designing edge weights based entirely on comprehensive sensitivity and its distribution ratio, an intuitive and accurate indicator is established for transmission path selection. Links with high security weights are prioritized for transmitting highly sensitive data, which helps to avoid the overall risk caused by single-point security failures. Furthermore, dynamic monitoring of latency and load factors optimizes bandwidth utilization efficiency, provides performance assurance for frequently used low-sensitivity data, and provides higher security path support for sensitive data.

[0115] The Dijkstra algorithm is used to traverse all nodes with the shortest path selection mechanism, and the edges connected to high-priority nodes are processed first. The terminal nodes are dynamically updated with the accumulated path weight each time, so that the data of each priority group can be transmitted along the path with the largest weight.

[0116] By eliminating the need for uniform resource allocation during transmission of data in different priority groups, highly sensitive data is prevented from being delayed or exposed to risks. Meanwhile, the path for transmitting low-sensitivity data at high frequencies is more inclined towards performance optimization.

[0117] S4, based on priority groups, classifies and determines candidate encryption algorithms for each priority group, and performs data fragmentation on the type data in the priority group;

[0118] Preferably, the encryption algorithm is graded based on priority groups, and candidate encryption algorithms are determined for each priority group. Furthermore, the data in each priority group is fragmented, including...

[0119] Priority is ranked based on the average comprehensive sensitivity score of priority groups. The priority threshold is set by the sum of the historical average score and twice the standard deviation. Priority groups with average scores greater than the priority threshold are marked as high priority groups, and priority groups with average scores less than or equal to the priority threshold are marked as low priority groups.

[0120] The sets of candidate encryption algorithms are determined based on the high-priority group, the low-priority group, and the high-latency link, where the high-priority group includes homomorphic encryption algorithms and the low-priority group includes the AES-256 algorithm.

[0121] The corresponding encryption algorithm is determined based on each edge link of the optimal path;

[0122] Data fragmentation is performed on the type data in the priority group, where the fragment size u is dynamically adjusted based on link load and bandwidth capacity, as shown below:

[0123] ;

[0124] in Indicates the size of the slice. This represents the minimum link bandwidth in the path. Indicates the total size of the transmitted data. Indicates the maximum number of supported shards;

[0125] Each fragment is encrypted according to the encryption algorithm of the path and its corresponding link.

[0126] By using homomorphic encryption for high-priority groups and AES-256 for low-priority groups, a flexible balance between security and performance is achieved. Homomorphic encryption for high-priority group data enables direct computation and processing while encrypted, protecting the privacy of highly sensitive data. Meanwhile, AES-256 for low-priority groups avoids excessive network performance overhead caused by high-priority encryption, ensuring efficient data transmission.

[0127] By combining the corresponding encryption algorithm for each side link of the optimal path, the optimization of transmission security and path dynamic adaptability is achieved. Each link is dynamically allocated its required combination of encryption algorithms along the path, so that different links can adaptively select encryption strategies according to their sensitivity and actual conditions, thereby strengthening the reasonable coupling between encryption strategies and path transmission conditions.

[0128] By using type data fragmentation in priority groups, lightweight and flexible data stream transmission is achieved. The fragment size is dynamically adjusted according to the link load and bandwidth capacity, avoiding links with low load in the path from becoming bottlenecks. This optimizes the utilization of transmission resources and reduces latency caused by link congestion or load imbalance during data transmission.

[0129] By combining fragment size with encryption algorithms to encrypt each fragment, distributed encryption control and fine-grained security management are achieved. The fragmentation mechanism, which dynamically adapts to link bandwidth and load, optimizes the algorithm to achieve a global balance between efficient transmission and security.

[0130] S5 performs storage verification on encrypted data fragments according to priority levels, constructs an LSTM model to predict link attributes, and performs anomaly analysis based on prediction errors.

[0131] Preferably, the encrypted data fragments are stored and verified according to priority levels, including:

[0132] Differentiated chain storage rules are designed for encrypted data sharding, including full encrypted storage and private key management for high-priority data, and partial data transmission through cross-chain transfer protocol for low-priority data, with block-level verification of integrity.

[0133] For nodes with different storage priorities, the integrity of the uploaded data fragment chain is verified step by step.

[0134] By using full encryption for high-priority encrypted data storage and managing it with private keys, strict privacy protection for highly sensitive data is achieved. By using a cross-chain transmission protocol to transfer part of the data for low-priority data and performing block-level verification, storage efficiency and integrity assurance for low-priority data are optimized. By verifying the integrity of the uploaded data fragment chain level by level for different priority nodes of stored data, efficient control of layered data storage security verification is achieved.

[0135] Furthermore, an LSTM model is constructed to predict link attributes, and anomaly analysis is performed based on the prediction error, including...

[0136] A deep learning model based on LSTM neural network is constructed, including an input layer, an LSTM layer and an output layer. The input layer takes in a time series consisting of link transmission delay, bandwidth utilization, link packet loss rate and retransmission count. The LSTM layer extracts the temporal relationship of the time series and outputs the predicted value for the next time step through the output layer.

[0137] The model is trained using calibrated training set data. The cross-entropy loss function is selected to calculate the difference between the predicted class probability and the actual label. The Adam optimizer is used for gradient descent optimization to update the model parameters. If the model loss no longer decreases significantly during continuous iteration, the iteration stops and the model parameters are output.

[0138] Collect real link status data, calculate the predicted Euclidean distance based on the model output, and use it as the prediction error. Use the sum of the mean of historical prediction error values ​​and twice the standard deviation as the detection threshold. If the prediction error is greater than or equal to the detection threshold, it is judged as data anomaly, and the path selection is re-performed.

[0139] By extracting temporal relationships based on LSTM neural networks and predicting the link state value at the next moment, accurate modeling of dynamic link behavior capture is achieved. By triggering path selection replanning when the prediction error is greater than or equal to the detection threshold in anomaly detection, efficient protection of the security of sensitive data transmission is achieved.

[0140] This embodiment also provides an artificial intelligence-based smartwatch data security transmission system, including:

[0141] The type data management module collects multi-source type data from smartwatch sensors and aggregates collection log information, including access logs and public match counts for each data type.

[0142] The comprehensive scoring module calculates a comprehensive sensitivity score for each type of data by combining privacy score, importance score, and real-time score.

[0143] The priority division module analyzes the distribution characteristics of the comprehensive sensitivity score, performs preliminary grouping based on the comprehensive sensitivity score, and optimizes the priority grouping and its group center value based on the distribution skewness coefficient.

[0144] The path planning module and the graph construction module apply Dijkstra's algorithm to select the path with the largest cumulative edge weight as the optimal data transmission path.

[0145] The hierarchical data storage module analyzes and encrypts data according to priority, performs data fragmentation, and stores data in a hierarchical manner.

[0146] The anomaly detection module constructs an LSTM neural network to extract temporal relationships and predict the link state attributes at the next time step, thereby performing anomaly detection.

[0147] This embodiment also provides a computer device applicable to the data security transmission method for smartwatches based on artificial intelligence, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the data security transmission method for smartwatches based on artificial intelligence as proposed in the above embodiment.

[0148] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0149] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the data security transmission method for a smartwatch based on artificial intelligence as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0150] In summary, this invention uses the entropy weighting method to dynamically adjust the weight of each parameter through information entropy, thereby enhancing its influence on the overall sensitivity score and reducing its influence on parameters with weaker discriminative ability. By introducing the distribution ratio of priority groups, it ensures that the node weights reflect the distribution characteristics of overall sensitivity while avoiding the meaningless amplification of the influence of certain secondary sensor data due to the diversity of a single device node. By avoiding a one-size-fits-all approach to equal resource allocation during the transmission of data from different priority groups, it avoids the lag or risk exposure of highly sensitive data. At the same time, the path for transmitting low-sensitivity data at high frequencies is more inclined towards performance optimization. By fragmenting data of different types in the priority groups, it achieves lightweight and flexible data stream transmission, with the fragment size dynamically adjusted according to the link load and bandwidth capacity.

[0151] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for secure data transmission in a smartwatch based on artificial intelligence, characterized in that, include: Collect multi-source data using smartwatch sensors to create a categorized dataset. The probability of disclosure is determined based on the number of accesses to each type of data in the access log and the number of public matches in the user-published database. A dynamic adjustment coefficient for potential sensitivity is calculated based on the total number of collections for each type of data. A privacy score is calculated based on the dynamic adjustment coefficient and the probability of disclosure. The business function coverage of each type of data is statistically analyzed, and an audit index is determined based on the number of user requests. An importance score for each type of data is calculated based on the coverage and the audit index. A real-time score is determined based on the data latency analysis. An entropy weight method is introduced to dynamically calculate the weights and calculate the comprehensive sensitivity score. The k-means clustering algorithm is used to initially group all comprehensive sensitivity scores to obtain several initial clusters. The elbow method is used to determine the number of initial clusters. The silhouette coefficient of each cluster is calculated. A dynamic balancing weight factor is introduced to optimize the silhouette coefficient and determine the number of optimized clusters. Based on the number of optimized clusters, clustering is performed again to obtain priority groups and the center value of each priority group. A graph model is constructed consisting of smartwatch devices and each data transmission node in their data transmission path. For each data transmission node, the global weight of the node is calculated based on the proportion of data in each priority group and the center value of each priority group. The edge weight of the link is calculated based on the transmission delay, maximum load, and global weights of the two ends of each link. With the goal of maximizing the total weight of the path, the Dijkstra algorithm is used to perform path selection decision and determine the optimal transmission path. The system is tiered based on priority groups, and candidate encryption algorithms are determined for each priority group. Data fragmentation is performed on the type data in the priority groups. The encrypted data is segmented and stored for verification according to priority levels. An LSTM model is constructed to predict link attributes, and anomaly analysis is performed based on the prediction error.

2. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 1, characterized in that: The introduction of the entropy weight method to dynamically calculate weights and calculate the comprehensive sensitivity score includes, The data access frequency data of the collected data types is statistically analyzed. The number of log accesses is determined based on the access logs of each data type. At the same time, the number of public matches of each data type is obtained from the user-published database, and the public probability is calculated. Analyze the dynamic adjustment coefficients for potential sensitivity based on the total number of times each type of data is collected; Privacy scores are calculated by combining potential sensitivity and public probability. In statistical data, the business function coverage of different sensor types is analyzed, and an audit index is defined by considering the number of user requests. Importance scores are then calculated. For each type of data The real-time performance score is determined by analyzing the delay between the timestamp data and the current time. The data is normalized based on its real-time score, importance score, and privacy score, and the entropy weight method is introduced to dynamically calculate the weights and calculate the comprehensive sensitivity score.

3. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 2, characterized in that: The global weight of the node is calculated based on the proportion of data in each priority group and the center value of each priority group. The edge weight of the link is calculated based on the transmission delay, maximum load and global weight of the two end nodes of each link. With the objective of maximizing the total path weight, Dijkstra's algorithm is used to perform path selection decisions and determine the optimal transmission path, including... Distribution analysis is performed on the comprehensive sensitivity scores of different types of data to calculate the standard deviation of the comprehensive sensitivity distribution, and the skewness of the data is determined based on the skewness coefficient. Based on the set of comprehensive sensitivity scores, the k-means clustering algorithm is used to initially group all comprehensive sensitivity scores, and the total within-group squared error is calculated. Based on the total within-group squared error and the initial group number, an elbow curve is plotted. The elbow point, i.e. the initial group number, is the position where the decrease in the total within-group squared error slows down significantly, and the value of the initial cluster is determined. Based on the clustering results, the weighted silhouette coefficient is verified, the silhouette coefficient is calculated, and the weighted optimization is performed to obtain the optimized centroid number. The comprehensive sensitivity score data is then clustered into priority groups, and the center value of each group is attached. Define smartwatch devices as device nodes, and set them as intermediate nodes and terminal nodes respectively according to the data transmission gateway and server. Calculate the distribution ratio based on the distribution of the type of data collected by the nodes in the priority group, and define the global weight of each node based on the clustered group data. Calculate the edge weights based on the node's transmission links, the link transmission latency, and the global weight of the maximum load. Set the optimal objective based on the edge weights, and select the path with the largest total weight as the optimal path; The Dijkstra algorithm is used to perform path selection decisions. The highest priority node is selected from the adjacent nodes of the initial node. All edges connected to the node are traversed. For each edge, the path weight of the target node is updated according to the cumulative edge weight. If the cumulative weight is larger, the path source of the target node is updated. The target node is added to the queue. When the terminal node is dequeued, the iteration backtracks the path and generates the optimal path.

4. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 3, characterized in that: The process involves prioritizing data into groups, determining candidate encryption algorithms for each priority group, and performing data fragmentation on the data within each priority group. include, Priority is ranked based on the average comprehensive sensitivity score of priority groups. The priority threshold is set by the sum of the historical average score and twice the standard deviation. Priority groups with average scores greater than the priority threshold are marked as high priority groups, and priority groups with average scores less than or equal to the priority threshold are marked as low priority groups. The sets of candidate encryption algorithms are determined based on the high-priority group, the low-priority group, and the high-latency link, respectively. The corresponding encryption algorithm is determined for each side link of the optimal path, and the data of the type in the priority group is processed by data fragmentation, where the fragment size u is dynamically adjusted based on the link load and bandwidth capacity. Each fragment is encrypted according to the encryption algorithm of the path and its corresponding link.

5. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 4, characterized in that: The storage verification of encrypted data fragments according to priority levels includes, Differentiated chain storage rules are designed for encrypted data fragmentation, including full-data encryption storage for high-priority data and private key management; For nodes with different storage priorities, the integrity of the uploaded data fragment chain is verified step by step.

6. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 5, characterized in that: The process involves constructing an LSTM model to predict link attributes and performing anomaly analysis based on the prediction error, including... A deep learning model based on LSTM neural network is constructed, including an input layer, an LSTM layer and an output layer. The input layer takes in a time series consisting of link transmission delay, bandwidth utilization, link packet loss rate and retransmission count. The LSTM layer extracts the temporal relationship of the time series and outputs the predicted value for the next time step through the output layer. The model is trained using calibrated training set data. The cross-entropy loss function is selected to calculate the difference between the predicted class probability and the actual label. The Adam optimizer is used for gradient descent optimization to update the model parameters. If the model loss no longer decreases significantly during continuous iteration, the iteration stops and the model parameters are output. Collect real link status data, calculate the predicted Euclidean distance based on the model output, and use it as the prediction error. Use the sum of the mean of historical prediction error values ​​and twice the standard deviation as the detection threshold. If the prediction error is greater than or equal to the detection threshold, it is judged as data anomaly, and the path selection is re-performed.

7. The method for secure data transmission in a smartwatch based on artificial intelligence as described in claim 1, characterized in that: The process involves collecting multi-source data through the smartwatch's sensors to form a type dataset, including: Multi-source data is collected through smartwatch sensors, and each type of data is combined into a type dataset, where each type of data... This includes sensor type, data value, and timestamp.

8. A smartwatch data security transmission system based on artificial intelligence, based on the smartwatch data security transmission method based on artificial intelligence as described in any one of claims 1 to 7, characterized in that: include, The type data management module collects multi-source type data from smartwatch sensors and aggregates collection log information, including access logs and public match counts for each data type. The comprehensive scoring module calculates a comprehensive sensitivity score for each type of data by combining privacy score, importance score, and real-time score. The priority division module analyzes the distribution characteristics of the comprehensive sensitivity score, performs preliminary grouping based on the comprehensive sensitivity score, and optimizes the priority grouping and its group center value based on the distribution skewness coefficient. The path planning module and the graph construction module apply Dijkstra's algorithm to select the path with the largest cumulative edge weight as the optimal data transmission path. The hierarchical data storage module analyzes and encrypts data according to priority, performs data fragmentation, and stores data in a hierarchical manner. The anomaly detection module constructs an LSTM neural network to extract temporal relationships and predict the link state attributes at the next time step, thereby performing anomaly detection.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the AI-based smartwatch data security transmission method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the AI-based smartwatch data security transmission method according to any one of claims 1 to 7.

Images